Regulatory T Cell Development in the Thymus

Adaptive immunity evolved as a powerful defense mechanism to eliminate foreign pathogens and eradicate transformed cells. This system relies on two chief capabilities: extensive repertoire diversity and the ability to discriminate “self” versus “nonself” (1). In T cells, diversity is derived from random rearrangements of the TCRα and TCRβ loci (2, 3). However, diversity comes at a cost, as some of these rearrangements will generate self-reactive T cells capable of initiating pathogenic immune responses. The thymus acts as a training ground for T cells and plays a role in ensuring a diverse, nonself-focused TCR repertoire capable of eliminating pathogens. The process of generating a diverse TCR repertoire also leads to the development of many autoreactive T cells. Many of these autoreactive T cells are eliminated via clonal deletion in the thymus. However, self-reactive T cells do escape clonal deletion and, when left uncontrolled, elicit detrimental autoimmune diseases. Although several mechanisms evolved to control autoimmune responses, a specialized subset of suppressor CD4⁺ T cells, termed regulatory T (T_reg) cells, plays a particularly important role in maintaining immune homeostasis.

Over the past 20 years tremendous progress has been made in the identification and understanding of T_reg cells. This relatively small population, ∼1% of developing CD4 single-positive thymocytes and ∼10–15% of CD4⁺ T cells in secondary lymphoid organs, is responsible for maintaining immune homeostasis and is crucial for survival (4–9). T_reg cells are an incredibly diverse population with regard to both TCR repertoire and function. T_reg cells regulate numerous physiologic processes, including maternal–fetal conflict (10–17), germ cell tolerance (18), stem cell differentiation in the skin (19), muscle repair (20), adipocyte homeostasis and function (21–25), and retinal inflammation (26). In addition, T_reg cells also regulate effector immune responses in disease states, such as germinal center reactions (27, 28), inhibit overzealous T cell responses during infection (29–34), enhance effector T cell differentiation and memory formation to pathogens (35–37), inhibit tumor immunity (38, 39), and promote tolerance to environmental and commensal Ags (40–42). The burden of regulating these diverse processes has led the field to propose two broad functional classes of T_reg cells, defined by their ontogeny: peripheral-derived T_reg (pT_reg) and thymic-derived T_reg (tT_reg) cells. In this review we focus on tT_reg cell development.

The thymus has been an organ of immense curiosity for immunologists for some time. Although initial thymectomy experiments failed to reveal immunological consequences (43), subsequent work revealed a central function in immune responses (44–46). Work as early as 1962 by Jacques Miller suggested a role in immune tolerance, as day 3–thymectomized (d3Tx) mice succumbed to an autoimmune wasting disease by 3 months of age (47). A seminal study in 1969 described that d3Tx mice but not day 7 or later- thymectomized mice developed autoimmunity of the ovary that could be rescued by a thymus transplant (48). Work by Gershon, Kondo, and colleagues (49–51) subsequently showed that thymocytes could produce dominant tolerance during immune responses to sheep RBCs and coined the term “suppressor T cells.” Together, these works suggested the existence of a population of thymus-derived suppressive T cells that had delayed kinetics of thymic export.

Although the concept of immune suppression was clearly correct, early models for explaining this process proved unsatisfactory. Most notably, it was suggested that suppressor T cells could function via a soluble factor encoded in the MHC locus I-J (52). However, the I-J locus was eventually found not to encode a unique protein (53). This led many to reject the concept of a unique population of T cells capable of immune suppression (54). Despite these controversies, work in the early 1980s already suggested the presence of a subpopulation of T cells, defined by anti–Lyt-1 (later described as CD5) Ab positivity, that were capable of suppressing autoimmunity in d3Tx mice (55). A seminal study by Sakaguchi et al. (4) in 1995 discovered that CD25⁺ T cells were necessary and sufficient for suppressing autoimmune responses. The identification of CD25 as a marker of suppressive T cells was critical to add legitimacy to the field (4). A follow-up study connected this concept to autoimmunity observed in d3Tx experiments, as d3Tx prevented accumulation of CD25⁺ cells in the periphery of mice. Transfer of CD25⁺ cells into d3Tx mice was able to rescue autoimmunity, whereas transfer of CD25-depleted splenocytes caused autoimmunity in athymic mice, revealing that thymically derived CD25⁺ T cells were critical controllers of autoimmunity (56). Groundbreaking studies in humans suffering from immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome, and in scurfy mice identified a critical role for the transcription factor FOXP3 in T_reg cells (6, 7, 57). This led to the generation of a series of reporter mice to track FOXP3 expression in live cells (58–60), enabling functional T_reg transfer experiments. Additionally, protocols were developed to detect intracellular FOXP3 by flow cytometry that enabled tracking and quantification of T_reg cells in non–reporter mice and humans (61). The identification of CD25 and FOXP3 as useful markers of T_reg cells led to an explosion of studies seeking to understand T_reg cell development and function.

The prevailing paradigm of thymic T_reg cell development involves a two-step process (62, 63). Step one is driven by strong TCR stimulation in developing CD4 single-positive thymocytes. This causes the upregulation of the high-affinity IL-2R, CD25, as well as TNFR superfamily (TNFRSF) members GITR, OX40 and TNFR2, thereby generating CD25⁺FOXP3⁻ T_reg cell progenitors (T_regP). The second step is driven by cytokine-dependent conversion of T_regP into mature T_reg cells via upregulation of FOXP3. These CD25⁺FOXP3⁺ cells are mature T_reg cells that emigrate from the thymus and mediate tolerance. More recent studies have implicated an alternative CD25⁻FOXP3^lo T_regP cell population (64); differentiation of these T_regP depends on the same two-step process (65). In this review, we focus on the mechanisms that drive T_reg cell development in the thymus and summarize current evidence on how the thymus shapes the T_reg repertoire and functions to maintain comprehensive immune tolerance.

Whether the tT_reg cell TCR repertoire is enriched in self-reactive TCRs was initially controversial. For example, one group found extensive overlap between TCRs in conventional T cells (T_conv) and T_reg cells and suggested that T_reg cells respond to non–self-antigens (66). Likewise, analysis of AND TCR transgenic mice observed that inducing Ag expression increased the T_reg cell proportion but not numbers in the thymus, suggesting that engagement of cognate self-antigen was not driving T_reg cell development (67). Nevertheless, other studies have provided evidence that the T_reg cell TCR repertoire is more self-reactive than its conventional counterpart and that acquisition of agonist TCR stimulation is important in T_reg cell development. This view originated from early experiments observing the presence of CD25⁺ cells in the thymus of wild-type (WT) mice but not those expressing a transgenic TCR specific for foreign Ag (68). This hypothesis was confirmed in later studies showing that TCR transgenics could drive thymic T_reg cell development only when the cognate Ag was also expressed in the thymus (69). Further, TCR sequencing experiments on mice with reduced TCR repertoires observed that T_reg TCRs are largely distinct from T_conv TCRs (70, 71), but overlap with TCRs expressed by pathogenic self-reactive T cells in Foxp3^−/− mice (72). In addition, a series of experiments observed that intraclonal competition for cognate Ag limits T_reg cell differentiation (73, 74), suggesting that interaction with Ag, presumably self-antigen, is important for T_reg cell development. Later work used TCR transgenics with various affinity for OVA and observed a linear relationship between TCR affinity and T_reg cell development (75). OVA-specific T_reg cells develop in RIP–mOVA thymi with TCRs spanning a broad 3 log–fold response range. Whereas lower-affinity TCRs can drive T_reg induction, TCR affinity and T_reg cell niche size are directly correlated with higher-affinity TCRs driving increased numbers of T_reg cells (75). Further, analysis of Nur77–GFP transgenic reporter mice, in which GFP is expressed coordinately with TCR signal strength, observed that T_reg cells were interacting more strongly with self-antigens (76). For example, lower proportions of TCR transgenic thymocytes in chimeric mice led to increased CD25⁺ cell proportions and a higher NUR77-GFP signal, confirming that developing T_reg cells compete for self-antigen during lineage commitment. TCR signal strength has also been related to the competency of developing T_regP cells to respond to low levels of intrathymic IL-2, suggesting another mechanism that would bias a T_reg cell repertoire toward self-reactivity (65). More recent studies have shown that intermediate dwell times for TCR–peptide/MHC complexes facilitate T_reg cell differentiation, whereas shorter dwell times preferentially drive positive selection, and longer dwell times lead to clonal deletion (77). This evidence collectively suggests that T_reg cell interaction with thymically presented Ag, at some elevated threshold (Fig. 1A), is necessary for initiating T_reg cell development.

Thymic selection is defined by a cellular dilemma: without the presence of specialized cell subsets, such as pancreatic β cells, how is the T cell repertoire pruned of reactivity to tissue-specific Ags (TSA) uniquely encoded by these cells? This led to the hypothesis that these specialized self-antigens were, in fact, expressed at some low level in the thymus, an idea first corroborated by human data correlating thymic insulin expression and susceptibility to the development of diabetes (78, 79). Subsequent work revealed evidence of broad “promiscuous” gene expression in the thymus and attributed medullary thymic epithelial cells (mTECs) with the sole ability to produce these TSAs (80). These studies also correlated expression of the transcriptional modulator autoimmune regulator (AIRE), a gene previously linked to polysymptomatic autoimmunity (81, 82), with the presence of TSA expression in mTECs. This supposition was confirmed in a set of groundbreaking experiments showing AIRE expression was necessary for tissue-specific gene expression in mTECs. Mice that lacked thymic expression of these TSAs had increased numbers of autoreactive T cells in peripheral lymphoid organs, which led to multiorgan immune destruction and generation of autoantibodies (83). Likewise, HEL-reactive TCR transgenic T cells underwent clonal deletion when HEL was expressed under the control of the rat insulin promoter, an AIRE responsive locus in mTEC. The proportion of CD25⁺ thymocytes increased; however, because there was no change in the absolute number of these cells, the authors dismissed a role for T_reg cell development to these Ags (84). These observations led to the hypothesis that the main role of AIRE in central tolerance was due to clonal deletion of tissue-specific effector T cells.

Although some controversy exists, numerous studies have now defined a role for AIRE-mediated ectopic Ag expression in mTECs in tT_reg cell development. Early studies in human patients with autoimmune polyendocrinopathy candidiasis and ectodermal dysplasia, a disorder caused by mutations in AIRE, documented a loss of T_reg cells and alterations in their TCR repertoire (85). Further, expression of hemagglutinin via the AIRE promoter in mice led to the development of hemagglutinin-specific T_reg cells, which was dependent on MHC class II (MHC-II) expression on mTECs (86). However, a follow-up study in AIRE–OVA mice produced a counterpoint to this hypothesis, as MHC-II knockdown on mTECs caused an increase in OVA-specific T_reg cell development (87). This finding suggested that low levels of high-affinity Ags drive tT_reg differentiation, whereas higher expression of these same Ags resulted in clonal deletion. In addition, another study observed AIRE-dependent prostate-reactive T_reg cell development in the thymus (88). Interestingly, analysis of the TCR repertoire of T_conv and T_reg cells in WT and Aire^−/− mice found that cells normally directed toward the T_reg cell lineage were instead found in the T_conv lineage in Aire^−/− mice (89), suggestive of T_reg cell agonist selection via AIRE-driven Ags. A similar phenomenon is observed in human patients harboring AIRE mutations in which TCRs normally found in T_reg cells are found in the T_conv compartment (90). In addition to AIRE, the transcription factor FEZF2 also regulates expression of TSA in the thymus. Fezf2^−/− mice also developed multiorgan autoimmunity, but the spectrum of organs targeted was distinct from Aire^−/− mice (91). Fezf2^−/− mice have fewer T_reg cells in the thymus and an altered TCR repertoire, reiterating a role for TSA expression in T_reg cell development. These results point to a crucial role for mTEC-derived TSA in central tolerance and T_reg cell development.

Recently, a distinct stromal cell involved in initiating type II mucosal immune responses, the tuft cell, has been identified in the thymus. Tuft cells were found to resemble mTEC and produce IL-25, a major inducer of IL-4 production (92, 93). Tuft cells contribute to the Hassall corpuscle, a structure in the thymus previously associated with T_reg cell generation in humans via licensing thymic dendritic cells (DCs) to produce CD80 and CD86 via TSLP stimulation (94). Interestingly, we observed that mice lacking the transcription factor POU2F3, which is required for tuft cell development, have reduced numbers of FOXP3^lo T_regP, suggesting that tuft cells can influence T_reg cell differentiation (95). Although the mechanism for this remains unclear, it may be due to IL-25 production or the expression of unique TSAs by tuft cells such as taste receptors (93).

The thymic DC compartment consists of conventional DCs, including SIRPα⁺ and CD8α⁺ DCs, and plasmacytoid DCs (pDCs) (96). Earlier studies suggested that DCs favor clonal deletion over T_reg cell development (86, 97). However, experiments using MHC-II^−/− bone marrow chimeras clearly implicated a role for bone marrow–derived DCs in both clonal deletion and T_reg cell induction (98). Other data using in vitro models of T_reg cell development also observed efficient T_reg generation by conventional DCs and to a lesser extent by pDCs (98–100). Although the role of DCs in T_reg development has become clearer, the Ags they present, required for inducing tolerance, remain blurry. This is due to the paradox that tolerance to AIRE-driven Ags are frequently dependent on DCs (101). Mechanistic insight into this paradox was revealed in studies documenting Ag transfer from AIRE-expressing mTEC to medullary DCs (102, 103). Interestingly, AIRE⁺ mTEC^hi cells produce the chemokine XCL1 that recruits thymic CD8α⁺ DCs to the medulla, and Xcl1^−/− mice exhibit defects in T_reg generation (104). CD8α⁺ DCs are the dominant cross-presenting thymic DC subtype; thus, in addition to producing intrathymic Ags (105), AIRE also mediates recruitment of APC populations to the thymic medulla required for efficient T_reg induction. Subsequent work used TCR sequencing and TCR transgenics derived from TCRs isolated from T_reg cells to determine the relative contributions of DCs and mTECs on central tolerance (105). This study observed that for some Ags, mTEC and DCs played nonredundant roles in T_reg cell differentiation and clonal deletion. However, for other Ags, mTEC and DCs played redundant roles in T_reg cell selection because of transfer of Ag from mTEC to DCs. Indeed, more recent studies using a prostate-reactive TCR transgenic observed that DCs were required to generate T_reg cells in the thymus, despite expression of the Ag being AIRE dependent (106). These experiments highlight the complex interconnections between thymic DCs and mTECs necessary for broad induction of Ag-specific thymic T_reg cells.

The contribution of SIRPα⁺ DC and pDC in T_reg cell polarization is particularly interesting, as these represent migratory DC populations capable of trafficking peripheral Ags to the thymus and inducing T_reg cell differentiation (96, 107, 108). pDCs also survey the gut via a CCR9-dependent mechanism (109), a chemokine receptor also required for pDC thymic localization and induction of central tolerance to peripheral Ags (110). This could represent a mechanism to transport gut-derived environmental or commensal Ags to the thymus. However, the contribution of endogenous peripheral self- or non–self-antigen trafficking to the thymus in T_reg development remains an open question.

The presence of nontransformed B cells in the thymus was observed more than 30 years ago (111). Early studies observed that B cell–deficient animals failed to delete Mtv-9–specific T cells, but reconstitution of these mice with B cells rescued this deletion (112). Further, in vitro studies observed efficient deletion of thymocytes by thymic but not splenic B cells (113). More recent studies have confirmed a role for thymic B cells in deletional tolerance to self-antigens (114, 115). For example, B cells induce clonal deletion of KRN autoreactive TCR transgenic T cells (116). The role of intrathymic B cells in T_reg cell development is less clear. The first evidence that thymic B cells affect tT_reg cell development came from the observation that BAFF-Tg mice had more tT_reg cells than WT mice because of an increase in thymic B cells (117). However, tT_reg cell development was decreased when thymic B cells were derived from hen egg lysozyme–specific transgenic B cells, suggesting that a broad, self-reactive B cell repertoire was required to promote tT_reg cell development (117). Using in vitro differentiation models, it was also observed that B cells isolated from the thymus were able to polarize CD4⁺ thymocytes to the T_reg cell lineage in a contact-, CD80/86-, and MHC-II–dependent manner (118). These experiments suggested that B cells increase the presence of CD25⁺ T_regP cells but do not facilitate the subsequent conversion of T_regP cells to mature T_reg cells.

T cells reactive to B cell–encoded proteins (such as Ig) are deleted by thymic B cells (119–121). There is some evidence that T_reg cells may also be generated to BCR Ags (120), although whether this happens in the thymus is unclear. In mouse and humans, AID and CD40L deficiency results in autoimmunity that correlates with a decrease in the proportion of T_reg cells (122). These studies, combined with observations that thymic B cells induce T_reg cell development in an MHC-II–dependent manner, suggest that thymic B cell–induced T_reg cell generation is critical for comprehensive immune homeostasis. Moreover, it was observed that self-antigens drive thymic B cell class-switching, which was required for inducing tolerance to self-antigens and was dependent on AID (123). A thymic B cell–licensing process has also been described, wherein interactions with T cell–derived CD40L increases Ag presentation on thymic B cells and induces AIRE expression on these B cells (124). This raises the possibility that thymic B cells have a parallel function to mTEC in producing TSA. However, it is still unclear what specificities of tT_reg cells are dependent on thymic B cells and whether interactions with thymic B cells preferentially promote T_reg cell development via CD25⁺ or FOXP3^lo T_regP cells.

Prior to the identification of CD25 as a marker for T_reg cells there were hints that IL-2R signaling was important for immune tolerance. In 1993, Il2^−/− mice were generated; these mice had increased numbers of activated T cells and developed colitis-like disease (125). Similar observations were made in Il2ra^−/− and Il2rb^−/− mice (126, 127). This was initially puzzling, as IL-2 is a known T cell growth factor. Subsequent studies revealed that expression of IL-2Rβ, specifically in the thymus, was sufficient to rescue the autoimmune phenotype observed in Il2rb^−/− mice, suggesting a role for IL-2R signaling during tT_reg development (128). These findings were questioned by studies showing development of CD25⁻FOXP3⁺ T_reg cells in Il2^−/− mice (129–131) and that transfer of T cells from Il2^−/− mice could protect against experimental autoimmune encephalomyelitis (132). However, further analysis observed that, whereas Il2^−/− mice do develop a small population of CD25⁻Foxp3⁺ T_reg cells, IL2Rβ^−/− mice have a larger block in T_reg cell development (131, 133). Further experiments observed that the IL-2Rβ binding cytokines IL-2 and IL-15 were the major inducers of T_reg cell development (131), although IL-7 had limited capacity to induce FOXP3 expression (134, 135). These latter findings reconciled previous reports of T_reg cell development in Il2^−/− mice, suggesting that in the absence of IL-2 other cytokines drive T_reg development, although not as efficiently as IL-2. Further, Stat5^−/− T cells are unable to differentiate into T_reg cells, whereas constitutive activation of STAT5 in STAT5b-CA transgenic mice led to a striking increase in T_reg cell differentiation (136, 137). Together, these findings confirm the critical role STAT5 plays in T_reg cell development.

Other γC cytokines have also been evaluated for their effect on T_reg cell development. IL-4 potently inhibits induced T_reg cell generation, and IL-4 blockade increased T_reg cell differentiation both in vitro and in vivo (138). Moreover, IL-4 is unable to induce STAT5 activation in CD25⁺ T_regP cells, and Il4ra^−/− mice show no obvious defect in T_reg cell generation in the thymus (134). However, more recent work has observed that IL-4 stimulation of FOXP3^lo T_regP maintains FOXP3 expression and upregulates CD25. Further, Itk^−/− mice, which exhibit elevated IL-4 production, exhibited an IL-4Rα–dependent increase in FOXP3^lo T_regP and mature T_reg cells. Consistent with this observation, BALB/c mice also have increased tT_reg cell production that is diminished on the Cd1d^−/− background (95), which eliminates NKT2 cells responsible for producing excess IL-4 (139, 140). Thus, IL-4 may function as a survival factor or provide a direct differentiation stimulus for FOXP3^lo T_regP. However, the mechanism by which IL-4 promotes tT_reg cell development and the significance of this pathway remain unclear.

The cellular sources of cytokines needed for tT_reg development remain incompletely understood. T cells and DCs represent the most likely cellular sources of IL-2 for tT_reg differentiation. Recent studies have observed that DC-derived IL-2 was particularly important for inducing T_reg cell development in ex vivo thymic slice models (141). These experiments suggested that DCs create a niche for efficient T_reg cell development by providing the antigenic stimulation for T_regP cell generation and the cytokine responsible for driving T_reg cell maturation. However, more recent work, using Il2^fl/fl mice crossed to T cell (Cd4-Cre)–, DC (Cd11c-Cre)–, or B cell (Cd79a-Cre)–specific CRE recombinases, observed that T cell–derived IL-2 is necessary and sufficient to drive tT_reg cell development (142). Further, autocrine production of IL-2 was not required for conversion of T_regP into mature T_reg cells. It remains unclear what subset of T cells is producing the intrathymic IL-2 necessary for T_reg cell development. FOXP3 blocks Il2 transcription (143), likely precluding FOXP3^lo T_regP as producers of IL-2. However, CD25⁺ T_regP may be competent to produce intrathymic IL-2, as these cells are receiving strong TCR stimulation. Alternatively, IL-2 may also be generated by activated recirculating T cells in the thymus (Fig. 1B). Future studies are necessary to pinpoint the specific cellular sources of IL-2 in tT_reg cell development.

Generation of IL-7 and IL-15–reporter mice has provided initial insight into the cellular players producing these cytokines in the thymus. Using IL-7–GFP knock-in mice, it was observed that IL-7 is present in both the thymic cortex and medulla. However, on a per cell basis, cortical thymic epithelial cells produced more IL-7 than mTECs (144). The lack of robust IL-7 production in the thymic medulla may explain the negligible effect of IL-7 on T_reg development (134). IL-15–CFP reporter mice produced the opposite result: IL-15 was preferentially found in the thymic medulla (145). Interestingly, IL-15 production was highest in mTEC^hi cells, the most robust Ag-presenting subset of mTECs defined by high expression of AIRE. More work is required to understand the cellular sources of IL-15 that may be contributing to tT_reg cell development.

Transcriptional regulation of Foxp3 and the broader T_reg epigenetic signature is essential for proper tT_reg cell development. Experiments to reverse engineer the T_reg cell transcriptional network, surprisingly, revealed a highly redundant system (146). It was revealed that FOXP3 alone was insufficient to drive the stable T_reg cell transcriptional landscape. However, FOXP3 plus any one of a quintet of other transcription factors (EOS, IRF4, SATB1, LEF1, or GATA1) was sufficient to solidify the T_reg cell transcriptional signature. Deletion of EOS or LEF1 had no effect on T_reg development by themselves (147, 148), whereas the effects of IRF4 or GATA1 deletion on T_reg development remain unstudied. However, subsequent studies observed a critical role for SATB1 in tT_reg cell development. SATB1 deletion at the CD4⁺CD8⁺ thymocyte stage prevented subsequent establishment of T_reg cell superenhancers and caused inefficient Foxp3 expression during later T_reg cell differentiation (149). Early work suggested that TCR stimulation also facilitates T_reg cell epigenetic signatures (150, 151). However, more recent experiments using an Il2ra mutant mouse provide evidence that IL-2 signaling is important for initiating the T_reg epigenetic signature (152). Specifically, SATB1 positioning throughout the genome was interrupted in developing T cells in Il2ra mutant mice. These results suggest that IL-2 signaling is also important for SATB1 to establish the T_reg epigenetic signature. Finally, deletion of the transcription factors NR4A1/3 almost completely blocks tT_reg generation (153, 154). Whether NR4A family members or other transcription factors act in concert with SATB1 to establish a permissive state prior to Foxp3 upregulation remains an open question.

Several studies have shown a crucial role for NF-κB activation in T_reg cell development. In particular, c-REL activation is required for T_reg cell development (155–158). c-REL but not NFκB1 activation downstream of CD28 is required for developing T cells to become CD25⁺ T_regP (156). However, FOXP3^lo T_regP are highly dependent on both c-REL and NFκB1 expression (95). Moreover, p65 (RELA)-deficient thymi also contain decreased amounts of CD25⁺ T_regP and mature T_reg cells (159). RELA and c-REL play partially redundant roles in maintaining T_reg cell transcriptional signature and homeostasis, although deletion of RELA resulted in a more severe autoimmune phenotype than deletion of c-REL (159). These findings suggest that NF-κB family members may also be important in locking in a stable T_reg cell phenotype, although the precise function of each NF-κB member during tT_reg development in establishing the T_reg cell transcriptional signature is still uncertain.

A key step in the development of tT_reg cells is stable upregulation of Foxp3. Much effort has focused on the factors and regulatory elements that control Foxp3 expression. Several conserved regulatory regions in the Foxp3 locus have been identified. These include the Foxp3 promoter, three intronic enhancers (Cns1-3) (158), and the Foxp3 pioneer enhancer element Cns0 (149). Cns0 is targeted by the transcriptional regulator SATB1 and acts to poise the Foxp3 locus for active transcription (149). Later, during T_reg cell selection, Cns3 acts as a pioneer regulatory element in the Foxp3 locus to drive de novo Foxp3 expression. This pioneer function is dependent on agonist TCR stimulation– and CD28-induced activation of c-REL and binding of c-REL to Cns3 (157, 158). c-REL targeting to the Foxp3 locus arranges an enhanceosome that includes several other transcription factors important for Foxp3 expression, including RELA, NFAT, SMAD, and CREB (160). Cns3^−/− T_reg cells are biased toward higher self-reactivity, suggesting that c-REL targeting of Cns3 is required to sensitize the Foxp3 locus to TCR stimulation (161). Additionally, Cns3^−/− thymi are devoid of the less self-reactive FOXP3^lo T_regP cell population (95). These experiments suggest that Cns3 evolved, in part, to expand the repertoire of T_reg cells. Interestingly, deletion of an Il2ra enhancer element CaRE4 (162), which has been linked to autoimmune single nucleotide polymorphisms in humans (163–165), causes a mild block in CD25⁺ T_regP and mature tT_reg development (95). Thus, regulatory regions inside the Foxp3 locus as well as those outside of Foxp3 are required for proper T_reg cell development. Future studies will need to identify other enhancer elements critical for tT_reg cell development and determine the specific role these enhancers play in generating the mature T_reg cell repertoire and transcriptome.

Studies of early T_reg cell ontogeny (58) illustrated that CD25 expression precedes FOXP3 expression and that the thymic CD4⁺CD25⁺ compartment is comprised of both FOXP3⁺ and FOXP3⁻ cells (166). This data provided the first hint that CD4⁺ CD25⁺ FOXP3⁻ thymocytes may represent cellular progenitors for mature CD25⁺ FOXP3⁺ T_reg cells. Subsequent studies illustrated that CD25⁺FOXP3⁻ thymocytes represent the direct cellular progenitors of mature T_reg cells (62, 63). These studies provided a two-step model of thymic T_reg cell differentiation (Fig. 1A). In step one, agonist TCR stimulation generates a CD25⁺ T_regP cell, whereas in step two IL-2/STAT5 converts CD25⁺ T_regP into mature T_reg cells. Later studies connected these two steps, finding that TCR signal strength correlated with expression of three TNFRSF members, GITR, OX40, and TNFR2, and signaling via these TNFRSF members renders developing T_regP cells much more sensitive to IL-2 (65). Thus, higher TCR self-reactivity imputes a selective advantage for developing T_regP by allowing these cells to compete more effectively for IL-2, thereby biasing the T_reg cell repertoire toward self-reactivity.

More recently, an alternative T_regP population was identified, defined by low FOXP3 and lack of detectable CD25 expression (FOXP3^lo T_regP). Initial reports demonstrated that FOXP3^lo T_regP cells efficiently develop into mature T_reg cells, either in vitro to high-dose IL-2 (200 U/ml) or in vivo in the periphery of mice. However, this paper also suggested that FOXP3 is normally a proapoptotic protein and must be counterbalanced by γC cytokine stimulation, such as IL-2, for T_regP to survive thymic selection (64). Despite the lack of CD25 expression, FOXP3^lo T_regP cells are able to differentiate into mature T_reg cells in response to low-dose IL-2 (0.2–1 U/ml) (65, 95) or intrathymic transfer (95, 167). Interestingly, in competitive intrathymic transfer experiments, CD25⁺ and FOXP3^lo T_regP both differentiated into mature T_reg cells at similar efficiencies; it remains unclear how FOXP3^lo T_regP are capable of such IL-2 sensitivity while lacking CD25 expression. CD25⁺ T_regP experience greater TCR stimulation, as measured by NUR77–GFP signal intensity, than FOXP3^lo T_regP during thymic selection (95, 167). The TCR repertoire of these two T_regP populations overlap significantly with mature T_reg cells but much less so with each other (95). These observations suggested that these were unique T_regP populations selected by distinct interactions with self-antigens and contributed unique TCRs to the mature T_reg cell repertoire. Remarkably, T_reg cells derived from CD25⁺ T_regP but not FOXP3^lo T_regP could protect mice from experimental autoimmune encephalomyelitis, whereas T_reg cells derived from FOXP3^lo T_regP were able to consistently suppress colitis. Collectively, these data provide an updated model of tT_reg cell development, in which both CD25⁺ and FOXP3^lo T_regP contribute quantitatively equivalently but qualitatively distinctly, to the mature T_reg cell repertoire.

Despite decades of research directed at understanding the development of tT_reg cells, many questions remain unanswered. Although two cellular progenitors have been described that contribute to the mature T_reg cell repertoire, the precursors to each of these populations have not been effectively described. Preliminary reports have identified a CD122⁺ GITR⁺CD25⁻FOXP3⁻ T_regP precursor that can give rise to CD25⁺ T_regP via a c-REL–dependent mechanism (168). However, whether this population also represents the precursors to FOXP3^lo T_regP remains unclear. Defining the signals and relevant Ags that commit T_reg cell development via either T_regP pathway will be important for understanding the role each pathway plays in immune tolerance.

Cytokine signaling is clearly required for T_reg cell generation. However, more nuanced effects of cytokines on T_reg selection remain poorly defined. CD25 can be expressed on thymic DCs and mTECs (142); however, it is unclear if CD25 trans presentation (169) occurs in the thymus and, if so, what affect this has on T_reg cell selection. The role of IL-4 is also unclear; mice of different background produce distinct amounts of IL-4 (95, 139, 140), which could influence the tT_reg cell TCR repertoire and, possibly, susceptibility to different types of autoimmunity. Further, certain subsets of thymic APCs produce different cytokines such as IL-2 from DCs (141) and IL-15 from mTECs (145). Future studies directed at understanding how distinct cytokines affect T_reg development will likely produce interesting insight into how cytokine stimulation affects T_reg cell repertoires.

Another mystery in T_reg cell development is how T_reg cells develop that enforce tolerance to transitory states such as inflammation, puberty, estrous, or distinct metabolic states. Certainly, for B cell immune responses, there is evidence of thymus-induced T_reg tolerance to Ig Ags (119–122), and loss of tuft cells leads to the development of anti–IL-25 Abs (93). Further, development of inflammation-specific T_reg cells has been observed in the thymus (170). Interestingly, testosterone levels regulate AIRE-mediated TSA production (171), which may explain resistance to various forms of autoimmunity in males. Prepubertal males and females have similar levels of testosterone (172); thus, any differences imposed by this hormone likely occur after puberty has initiated in humans. The broad specificity of tT_reg cells needed to provide tolerance in transitory states is still poorly understood.

The tT_reg pool is composed of recently differentiated cells but also T_reg cells that have been retained following development (resident) or have recirculated to the thymus from the periphery (173–175). Studies with Rag2–GFP mice demonstrate that older GFP⁻ T_reg cells progressively accumulate in the thymus as mice age and represent the majority of thymic T_reg cells by ∼8 weeks of age (175, 176). However, the origin of these T_reg cells is debated, with some suggesting that they are mostly resident cells that never left the thymus (173) and others proposing that they are primarily recirculating cells (175). It has been difficult to distinguish between these two populations to determine their relative contributions to the tT_reg cell pool. Thymus transplantation studies demonstrate that T_reg cells migrate from the periphery to the thymus preferentially by comparison with T_conv (177). Additionally, mature RAG2–GFP⁻ T_reg cells in the thymus have a similar gene expression profile with splenic T_reg cells, and their TCR repertoire shows evidence of peripheral modification, supporting the possibility that these cells are recirculating (175). Resident and recirculating T_reg cells have been shown to compete with developing thymic T_reg cells for access to IL-2 and limit their differentiation to the T_reg cell lineage (141, 175). The immunological benefit of restricting new T_reg cell development is unclear. It is possible that these older T_reg cells also compete with thymocytes for Ag, costimulatory ligands, and TNFRSF ligands necessary for T_reg cell development. The presence of a large population of recirculating or resident T_reg cells represents both an opportunity to understand the biological importance of these recirculating T_reg cells and a problem, as these RAG2–GFP⁻ T_reg cells contaminate analysis of de novo T_reg cell development. Cellular phenotypes for “old” contaminating T_reg cells have been proposed, including CCR6⁺CCR7⁻ (177) as well as CD73⁺ (95); these markers should be used to exclude old T_reg cells in studies of de novo tT_reg development.

Finally, despite years of debate, controversy still exists over the relative role of tT_reg cells and pT_reg. The hypothetical requirement for pT_reg is at mucosal surfaces (178), where diverse Ags are being surveyed, or during pregnancy, in which ectopic alloantigens are contributed by the male gamete (10). Several studies suggest a role for thymic deletion and T_reg cell selection in mucosal tolerance (66, 179–181), whereas other studies argue for the importance of pT_reg generation (40, 178, 182–185). More recent studies have suggested that some populations of thymic T_reg cells are required to polarize T_conv to pT_reg, perhaps relating these disparate findings (182, 186). Likewise, T_reg cells derived from thymic FOXP3^lo T_regP were able to suppress colitis, suggesting tolerance to commensal organisms can be induced by specific tT_reg cell subsets (95). Further experimentation is required to conclusively delineate the unique and overlapping responsibilities of pT_reg and tT_reg cells in immune tolerance.

The evolutionary constraints placed on T cell selection in the thymus are immense; exogenous pressure from pathogens places a high priority on TCR diversity, whereas endogenous pressure requires removal of self-reactive and potentially pathogenic T cells. Thus, T_reg cell development represents a mechanism that allows this leaky selection system to persist and focus effector T cell responses on non–self-antigens. Future studies defining endogenous T_reg cell antigenic targets and the thymic populations required to produce these Ags will be required to understand the complex processes that govern the selection of a competent repertoire of tT_reg cells. Further, understanding the role of Ag specificity of T_reg cells in homeostatic, inflammatory, or autoimmune contexts will be crucial in linking thymic selection to peripheral homeostasis.

We thank Dalton Hermans for critical comments on the manuscript.